This invention relates to a multiprocessor system comprising a plurality of processor units which are loosely coupled to one another.
A conventional loosely coupled multiprocessor system of the type described is disclosed by James A. Katzman in "A Fault-Tolerant Computing System" published on January, 1979 (first revision), by Tandem Computers Inc., Calif. The conventional loosely coupled multiprocessor system comprises a plurality of processor units having independent operating systems and a plurality of peripheral devices used by the processor units in common. At any rate, each of the processor units can individually carry out a processing operation by the use of a set of control signals.
When one of the processor unit falls into disorder as a result of occurrence of a fault, that processor unit interrupts the processing operation thereof. The processor unit may be called a faulty processor unit. Under the circumstances, another of the processor units, namely, a normal processor unit can take over the processing operation interrupted by the faulty processor unit. Thus, the normal processor unit gives relief to the fault of the faulty processor unit.
In order to make another processor unit take over the processing operation of the faulty processor unit, the control signal set should be transferred from the faulty processor unit to the normal processor unit. A transfer circuit may be placed between the processor units for use in transferring the control signal sets therebetween. Inasmuch as a great amount of control signals should be transferred through the transfer circuit, it takes a long time to transfer all of the control signal set from the faulty processor unit to the normal processor unit. As a result, each processor unit is subjected to a heavy overload. Use of the transfer circuit is therefore not practicable.
A method has practically been adopted whereby transfer of the control signal set is carried out during an intermission predetermined in each processing operation. With this method, the processing operation of the normal processor unit should be returned back to the predetermined intermission and reexecuted when one of the processor units falls into disorder. Therefore, the normal processor unit can momentarily take over the processing operation of the faulty processor unit.